| Author |
 Topic  |
|
|
smhayes
New Member
21 Posts |
 Posted - Jun 04 2007 : 4:54:13 PM
|
Has anyone found values for the intensity arguments in BET that work well for the SPGR images off the 3T? I've been playing w/ different values, and I am having difficulty getting a nice skull-stripped volume. I'm using SCIC processed images, although there is an obvious posterior to anterior decrease in signal intensity across the brain, which is likely contributing to my difficulty finding appropriate thresholds. If someone has found values that work well, I'd appreciate a quick email. My attempts to systematically manipulate the threshold values haven't worked yet.
Thanks,
Scott Hayes |
|
|
petty
BIAC Staff
USA
453 Posts |
Posted - Jun 05 2007 : 09:37:53 AM
|
from the command line i've used these values/commands and have never had a problem:
bet INPUT OUTPUT -f 0.5 -g 0
what anat series number are you using (ie: 002 or 200)? all of my scic images have looked pretty uniform. |
 |
|
|
smhayes
New Member
21 Posts |
Posted - Jun 05 2007 : 10:19:59 AM
|
I'm using a "100" series.
I've tried the BET defaults, and numerous variations. According to FSL archives, inputting centerpoints should help, but when I include those arguments, it doesn't output a volume (but that's a separate issue).
Maybe my expectations are too high? Do your stripped brains contain non-brain matter? I'm expecting a nicely stripped brain that doesn't include non-brain matter (e.g., around the brainstem, eyes, etc). However, based on the parameters I've been trying, that doesn't seem possible, at least w/o some manual segmentation. Presumably the assumption is that some non-brain matter doesn't significantly impact registration? I'm processing a data set now to check the registration. |
|
|
|
petty
BIAC Staff
USA
453 Posts |
Posted - Jun 05 2007 : 10:53:13 AM
|
isn't 100 typically the scic'd sagital localizer?
here's an example of mine (from the 4T so not as clear though)
 |
|
|
|
smhayes
New Member
21 Posts |
Posted - Jun 05 2007 : 11:28:49 AM
|
Can you email me your image directly? I can't view it from the forums.
I meant it's a hundred series, indicating that it was SCIC'd--the actual number is 500, but that doesn't matter.
|
|
|
| |
Topic |
|
FSL BET gradient thresholds
